Optimisation of chemical batch plant design: mathematical models and algorithms for strategic decision support
14 June 2019
University of Antwerp, City Campus, Hof van Liere, Frederik de Tassiszaal - Prinsstraat 13 - 2000 Antwerp (route: UAntwerpen, Stadscampus
Prof. dr. Trijntje Cornelissens, Prof. dr. Johan Springael
PhD defence Floor Verbiest - Faculty of Business and Economics
Food additives, pharmaceuticals and lubricants are, among other products, typically manufactured in chemical batch plants. The construction of such batch plants comes with significant investments, which require appropriate capacity assessments. Such assessments, or strategic capacity decisions, are the main topic of the strategic batch plant design problem. This problem entails determining the optimal number and size of equipment units for every production stage, as well as the optimal operational planning guidelines, so as to minimise total costs while satisfying both demand and design related constraints. These demand and design constraints generally state that the designed plant should be large enough to produce a given, total demand for a range of products within a given production horizon.
The batch plant design problem has been studied extensively in literature over the past four decades and is usually formulated as a mixed-integer linear programming problem that is solved exactly. There is, however, still a need for bringing the existing models and solution techniques a step closer to reality. With this goal in mind, a first contribution of this dissertation relates to the modelling of real-life chemical batch plants. Such plants typically consist of parallel production lines, i.e. lines that can be dedicated to particular products or product families, which we included as a design option into existing plant design models found in literature. Secondly, we investigated the integration of the operational use, i.e. how the plant is used, into the plant design problem with parallel production lines. Apart from a single product campaign mode of operation used earlier, two additional, more complex, modes were examined: mixed-product campaigns and network planning.
As a third contribution, different solution techniques, such as a matheuristic optimisation algorithm, were developed to solve these design problems with parallel production lines within limited computation time. Finally, we considered the multiperiod batch plant design problem and proposed a multiperiod delivery framework. This framework allows to systematically combine different delivery and production planning characteristics that arise in such a multiperiod context. We performed an exploratory study to investigate the impact of this multiperiod delivery framework on the plant design, and this for both a plant operating in a single and a mixed-product campaign mode.